JP2993120B2 - Method and apparatus for canceling random FM noise - Google Patents

Description

Description: TECHNICAL FIELD The present invention relates generally to FM (frequency modulation) wireless communication systems, and more particularly to FM wireless systems based on digital signal processing (DSP) technology, and more particularly to transmissions induced by fading transmission channels. The present invention relates to a method for canceling random FM noise generated by undesired phase modulation of a signal.

BACKGROUND OF THE INVENTION In order to gain a better understanding of the problems associated with the present invention, Rayleigh fading
You have to understand the concept. Rayleigh fading is a rapid change in the magnitude and phase of a received signal due to multipath propagation. A signal that has undergone this Rayleigh fading becomes extremely unpleasant to the listener if the magnitude of the fading is large enough to cause the amplitude of the received signal to drop substantially and instantaneously. As a result, the listener will hear an unpleasant noise burst.

Rayleigh fading frequently occurs when a mobile wireless user travels on a road and passes through relatively close, nearby objects, such as telephone poles, buildings, bridges, and the like.
Variations in electric field strength occur in the mobile antenna of the vehicle. Such a change in the electric field intensity causes a variation corresponding to the envelope of the received signal. Unpleasant noise bursts and pops in the received speech that substantially match the minimum envelope are characterized as random FM noise.

Conventional methods of improving the response to Rayleigh fading signals are mostly analog, requiring valuable space in mobile radio equipment and increasing cost. Moreover, most of these methods are double-ended, requiring some signal processing to be added to both the mobile device and the base station. One method for improving the performance of a receiver in a Rayleigh fading environment described in U.S. patent application Ser. No. 091,160, filed Aug. 28, 1987, and assigned to the same assignee as the present invention, is to use a fading event. While the receiver audio is actually muted. Although this method substantially eliminates noise bursts and pops, it can lose information and become meaningless. Accordingly, the method described in the above-cited patent application performs a mixing operation (multiplication of two or more signals), resulting in a tone-encoded squelch (tone-
coded squelch or slow trunking data (low
Audible (sub-audibl) such as -speed trunking data
The superimposed signal in e) may be converted to an audible component of the spectrum, causing unpleasant rumble in the received voice.

Therefore, there is a need for a method for canceling random FM noise using a modern DSP (Digital Signal Processor) in order to save space and reduce the cost of analog components. There is also a need for a way to address the effects of fading events without muting the audio. In addition, random
There is a need for a method that negates the effects of FM noise and prevents audible signals from being heard in the received voice.

SUMMARY OF THE INVENTION In accordance with the present invention, there is provided a method for canceling random FM noise that substantially eliminates discriminator output noise pulses. Cancellation is performed by processing the envelope of the FM modulated signal and determining characteristics indicating the waveform, amplitude and generation time of the noise pulse. The discriminator output signal is processed to determine a characteristic indicative of a noise pulse direction. A counter pulse is generated based on these characteristics, and the counter pulse and the discriminator output are combined,
Substantially cancels random FM noise.

BRIEF DESCRIPTION OF THE DRAWINGS FIG. 1 shows the real and imaginary parts of a two-waveform fading parameter and the generated amplitude.

FIG. 2 shows the real and imaginary parts of the Rayleigh fading parameter and the generated amplitude.

FIG. 3 shows the normalized derivative of the fading envelope centered on the minimum envelope.

FIG. 4 shows the relationship between the normalized envelope derivative and the Rayleigh fading discriminator output.

FIG. 5 is a block diagram of the random FM noise cancellation algorithm of the present invention.

FIG. 6 shows the counter pulse and the noise pulse correctly aligned for the summation with noise pulse cancellation.

DETAILED DESCRIPTION OF THE INVENTION A faded signal can be modeled by:

f (t) = r (t) s (t) where s (t) = ^{ejθ (t)} is a desired FM signal, and r (t) = ρ (t) ^{ejψ (t)} is a fading signal.
Parameter.

 Therefore, the received signal can be expressed as follows.

f (t) = ρ (t) ej ^{[θ (t) + ψ (t)] When} this signal is applied to the FM limiter / discriminator, the generated output is:

d (t) = θ ′ (t) + ψ ′ (t) where Is the desired message signal, Is a random FM noise signal. During deep fades, P (t) contributes pulsed noise to the desired signal. These pulses have a significant amplitude for the desired signal and have a speech bandwidth (300
When containing energy at ~ 3000 Hz), these pulses degrade speech quality considerably.

The random FM noise cancellation method of the present invention,
The information contained in the fading envelope (ρ (t)) is used to substantially cancel random FM noise pulses at the discriminator output. The amplitude and waveform of the pulse are obtained from the envelope, and the direction (sign) of the pulse is obtained by appropriately filtering the discriminator output signal.

The effect of the present invention is based on the assumption that the real and imaginary parts of the fading parameter are approximately linear during the noise pulse. The fading parameter is
It was previously defined in exponential form as:

r (t) = ρ (t) ^{ejψ (t) In} rectangular form, the fading parameter can be expressed as:

r (t) = x (t) + jy (t) During the period of the random FM noise pulse P (t), r
The real and imaginary parts of (t) are approximately linear. That _{x (t) = m x t} + b x y (t) = m y t + b y this assumption can be completely a reasonable justification for two reasons. First, Rayleigh (Rician, multipath, etc.) fading is a low-pass process. Thus, during a time period substantially less than the minimum time period of the process, the signal is substantially linear. Unwanted noise pulses caused by fading are:
It has a much smaller width than the minimum period of this process. This linearity assumption is particularly true during the minimum envelope during which the noise pulse is generated.

FIG. 1 shows the real part (102) and imaginary part (103) of the two-waveform fading parameter, as well as the generated amplitude (101). Dual waveform fading consists of the direct reception of the desired signal and another reflected signal, the amplitude of both signals being kept constant. As FIG. 1 shows, the real part (102) and the imaginary part (103) of the fading parameter are:
It is the most linear at the minimum envelope.

FIG. 2 shows the real part (202) and imaginary part (203) of the Rayleigh fading parameter and the generated amplitude (201). Although the effect is less pronounced with the Rayleigh fading parameter than with two-waveform fading, the real part (202) and the imaginary part (203) are most likely to be linear at the minimum envelope as shown in this figure.

Based on this linearity assumption, the amplitude and waveform of the random FM noise pulse at the discriminator output are
It can be seen that the calculation can be performed using only the information included in the envelope. While this result is generally correct, the derivative shown below is for the case of simple linearity to minimize algebraic complexity.

The fading parameters could be expressed in rectangular form as:

r (t) = x (t) + jy (t) Therefore, the amplitude and phase of the fading parameter are given by:

Here, the following fading is assumed.

x (t) = b (where b is a constant other than 0) y (t) = mt (where m is a constant gradient other than 0) When r (t) is input to a limiter / discriminator for FM demodulation, The output is given by:

P (t) becomes maximum when t = tm = 0, that is, when ρ ² (t) is the minimum value. Therefore, P _max can be defined as:

The fading envelope is Here, the derivative ρ ′ (t) of ρ (t) with respect to time is given by the following equation.

The normalized derivative of ρ (t) is defined as:

FIG. 3 is a plot of g (t) centered on the minimum envelope.

It can be seen that the magnitude of the random FM noise signal (| P _max |) is the peak-to-peak amplitude of the normalized derivative. _{That, | P max | = g max} -g from _min Figure 3, a _{g max = g (t max)} , g min = g
(T _min ). Since t _max and t _min are obtained by finding the solution of g ′ (t) = 0, the maximum and minimum times can be obtained using basic calculation methods. That is, Since ρ (t) is defined to never be zero, the solution to this equation is given by:

m ⁴ t ² = m ² b ² or Therefore, By substituting these values for g (t), g _max and g _min are obtained.

Finally, This relation also applies to the general case.

for the case of _{x (t) = m x t} + b x y (t) = m y t + b y Rayleigh fading, Figure 4 shows a noise pulse in the discriminator output by the dotted line, also a derivative of the normalized envelope by a solid line . It is clear from this figure that the magnitude of the noise pulse is equal to the maximum and minimum amplitude values of the normalized derivative.

The signal at the output of the FM limiter / discriminator in response to the above general linear fading is (Where k is a constant).

 It has already been demonstrated that the following equation holds.

Similarly to | P _max |, since ρ (t) is known, | k | can be obtained by the following equation.

Thus, the absolute value of the entire noise pulse at the discriminator output can be reproduced in both amplitude and waveform, using only the information contained within the fading envelope.

To substantially cancel the noise pulse at the discriminator output, only the direction of the noise pulse need be determined. The direction of the pulse can be detected by appropriately filtering the discriminator output signal. In the absence of audible signals, such as coded squelch or slow trunking data, the pulse direction can be detected via a low pass filter. When an inaudible signal is present, a properly designed low-pass or band-pass filter performs the function of determining the pulse direction. In some cases (ie, when the bandwidth of the message signal is low-pass), a high-pass filter can be used to determine the pulse direction.

Referring to FIG. 5, a block diagram of a preferred configuration of the method of the present invention, generally designated by the reference numeral 500, is shown. Since this method is designed for digital radio equipment, the noise cancellation algorithm is DSP, preferably M
It is configured using a DSP56000 device manufactured by otorola.

The algorithm receives an input from a digital IF filter in the form of a series of complex samples at the input of the discriminator (501). The algorithm necessarily runs at the discriminator sampling rate.

The size of each complex sample in the discriminator is
Calculated at 502. Samples of these sizes are the raw envelope (ρ) of the received signal.
Make This envelope signal is filtered by an envelope low pass filter (LPF) (503). The envelope LPF (503) is necessary for reducing the influence of noise on the envelope and the influence of ripple due to the IF filter. The filter introduces a delay into the calculations performed by the algorithm, so that if the introduced delay is an integer sample period, the envelope LPF (50
3) is preferably a linear phase and symmetric. Limiting the filter delay to an integral number of sample periods simplifies aligning the generated counter pulse with the actual noise pulse. This will be described in detail below. The envelope LPF is implemented in the DSP as a symmetric finite impulse response (FIR) filter with an odd number of taps.

In Figure 5, the envelope is marked as [rho ₁ LP
The output of F (503) is processed via two parallel paths. In the lower path, the derivative of ρ ₁ is
Calculated in 4. The expression that appears in block 504, 1-
z ⁻¹ is one of the possible z-transform approximations of the derivative, as is well known in the art, and introduces a delay of 標本 the sample period.

To compensate for this 1/2 sample period delay, the upper path includes a two tap filter (505) that adds the 1/2 sample period delay. The equation [1 + z ⁻¹ ] / 2 is one possible z-transform of a suitable delay process. The delayed envelope is subjected to an inversion process (506), and the envelope derivative and the inverted envelope are multiplied at block 507 to become a normalized derivative.

The sign of the counter pulse is determined by a pulse sign filter using the output of the discriminator (501) as an input.
ter) (508). This filter covers the frequency domain whose passband contains the minimum energy from normal speech and the audible signal, and furthermore the noise pulse
Must be designed to maximize energy. Since the random FM noise pulse is low pass in nature, the pass band of the code filter must be in the frequency range below the normal speech cutoff frequency of 300 Hz.

To minimize noise-to-pulse interference, the impulse response of the code filter is short and the main lobe
There must be only one. Noise pulses that are so severe that they require cancellation become substantially impulse to the code filter. Therefore, the delay of the code estimation from the detected pulse center is the maximum amplitude excursion of the impulse response of the filter.
rsion) delay.

When the normalized derivative calculated in block 507 of FIG. 5 passes the pulse detection threshold of FIG. 3, it causes the algorithm to begin collecting data to generate a counter pulse. In a preferred embodiment, data collection is initiated by a normalized derivative passing a minimum threshold, but by a normalized derivative passing a maximum threshold or the absolute value of a normalized derivative exceeding a predetermined threshold. You can also. The pulse detection measurement block (509) detects this threshold crossing and determines the local minimum g between the threshold crossing and zero crossing.
(T) Start searching for (g _min ). Next, the local maximum value (g _max ) is determined, and the absolute value of the pulse amplitude is calculated as described above. The output of the pulse code filter (508) performs code estimation.

Here, only the pulse waveform remains. As described above for the derivative in the general case, the pulse waveform is the ratio of the envelope value at the pulse center (ρ (t _m )) to the envelope amplitude around the fading event (ρ (t)). Determined by the square. The value of the envelope amplitude at the center of the pulse is
From the delayed envelope passed to the pulse detection measurement block (509). The determination of the final pulse shape is made by multiplying by the appropriately (τ _i ) delayed samples of the inversion envelope provided by the delay line (511), the pulse generator block (510).
It is performed in. Pulse detection measurement block (509)
Provides the necessary information about the pulse amplitude and sign to the pulse generator (510). Since the algorithm is implemented in the DSP, the delays introduced by the individual processes are known exactly, so τ _i is determined and the inverted envelope information is used so that the counter pulse matches the noise pulse at the discriminator output. Is easily delayed by an appropriate amount.

Output, the length tau _c delay line discriminator (501) (51
2) is also entered. Here, τ _c represents a cancellation delay. This cancellation delay represents the minimum speech delay required for the algorithm to work properly and must also be known exactly. The noise pulse at the output of the delay line (512) and the properly aligned counter pulse from the pulse generator (510) are added in an adder (513), and then a simple de-emphasis filter. (51
4) is processed and becomes the final audio output. Sixth
The figure shows the counter pulse (solid line) and the noise pulse (dotted line) correctly aligned to cancel the noise.

Since the correct operation of the noise cancellation algorithm of the present invention depends heavily on accurate characterization of the envelope of the received signal, it may be necessary to modify the algorithm slightly under operating conditions that tend to corrupt the envelope data. is there. These bad operating conditions include low S / N (s
ignal-to-noise ratio, adjacent channel or co-channel interference and frequency selective fading (frequenc
y selective fading).

Under good operating conditions, the normalized derivative has only one zero crossing during the acquisition period, as shown in FIG. But,
At low S / N, the normalized derivative is a function of noise as well as the desired envelope signal. Under these conditions, the normalized derivative has much more zero crossings. Another extraneous zero-crossing event is also caused by high frequency envelope ripple due to interference or frequency selective fading. Thus, whenever a second zero crossing occurs within a predetermined time after the first zero crossing, data collection is turned off and the algorithm is reset.

Computer simulations of Rayleigh fading have shown that random FM noise pulses never have an amplitude greater than 2 radians. Thus, if | g _max | or | g _min | exceeds the value 1.0, something is probably abnormal. In response to this error condition,
The pulse amplitude estimate obtained by the algorithm is set to zero, and a zero amplitude counter pulse is added to the discriminator output.

A third possible error condition may occur during the calculation of the counter pulse waveform. As mentioned above, this waveform is the square of the ratio of the minimum envelope to the amplitude of the envelope sample centered on the noise pulse event. This ratio must always be less than or equal to one. This is because the minimum envelope itself is a local minimum of the envelope sample amplitude centered on the noise pulse event. If this ratio is greater than one, an error has occurred and the pulse amplitude estimate is set to zero, preventing non-zero counter pulses from being added to the discriminator output.

Continuation of front page (58) Field surveyed (Int.Cl. ⁶ , DB name) H04B 1/10

Claims (10)

(57) [Claims] An FM-modulated RF signal having an envelope has a randomizer that substantially eliminates a discriminator output noise pulse.
An FM noise cancellation method comprising: (a) processing an envelope of an FM modulated signal to ascertain characteristics indicative of a noise pulse waveform, amplitude and generation time; (b) processing a discriminator output signal. Grasping at least one characteristic indicating the direction of the noise pulse; (c) generating a counter pulse based on the characteristic determined in steps (a) and (b); and (d) counter pulse. Synthesizing the discriminator output to substantially cancel the random FM noise pulse. 2. The step of processing the envelope of the FM modulated signal comprises the steps of: (a1) differentiating the envelope to provide an envelope derivative; (a2) dividing the envelope derivative by the envelope;
Obtaining a normalized derivative; (a3) calculating a maximum / minimum amplitude of the normalized derivative to give a noise pulse amplitude; (a4) calculating a minimum envelope generation time and calculating a noise pulse generation time (A5) the envelope amplitude at the minimum envelope;
2. The method according to claim 1, further comprising: calculating a square of a ratio of an envelope amplitude centered on the minimum envelope to an envelope amplitude to provide a noise pulse waveform. 3. The method according to claim 2, wherein the processing steps (a3) to (a5) are performed only when the magnitude of the normalized derivative passes a predetermined threshold. 4. The method according to claim 2, wherein the processing steps (a3) to (a5) are terminated if the normalized derivative has more than one zero crossing within a predetermined time. . 5. The step (b) of processing the discriminator output signal includes: (b1) processing the discriminator output signal to provide a filtered signal; and (b2) sampling the filtered signal to obtain a noise pulse direction. 2. The method of claim 1, further comprising: providing. 6. The method of claim 5, wherein the filtered signal is sampled at a time substantially corresponding to a maximum amplitude excursion of the filter's impulse response. 7. The method of claim 1, wherein the step of combining the counter pulse and the discriminator output comprises adding the counter pulse and the discriminator output. 8. A randomizer for substantially eliminating discriminator output noise pulses for an FM modulated RF signal having an envelope.
Means for processing an envelope of an FM modulated signal to ascertain a first characteristic indicative of a waveform, amplitude and generation time of a noise pulse; processing an output signal of a discriminator to generate a noise pulse Means for ascertaining at least one second characteristic indicative of a direction; means for generating a counter pulse based on the first and second characteristics; and combining the counter pulse and the discriminator output to produce a random number.
Means for substantially canceling the FM noise pulse. 9. The means for processing the envelope of the FM modulated signal includes: means for differentiating the envelope to provide an envelope derivative; means for dividing the envelope derivative by the envelope to provide a normalized derivative; Means for calculating the maximum and minimum amplitudes of the function and providing a noise pulse amplitude; means for calculating the minimum envelope generation time and providing the noise pulse generation time; an envelope amplitude in the minimum envelope and an envelope centered on the minimum envelope 9. The apparatus according to claim 8, further comprising: means for calculating a square of a ratio with the amplitude to provide a noise pulse waveform. 10. The means for processing the discriminator output signal further comprises: means for filtering the discriminator output signal to provide a filtered signal; means for sampling the filtered signal to provide a noise pulse direction. The method of claim 8, wherein